SRE Weekly Issue #541
A message from our sponsor, incident.io:
Your agent is only as useful as the tools it can reach, and only as safe as the limits you set.
Join incident.io’s engineering team for an Extensions showcase. We'll build a skill live, watch it run, and safely deploy.
Archaeological layers of P0 incident action items? There's a better way.
Brent Chapman
Wow, the original for this article was one I included all the way back in issue #16. I think I've quoted the concepts from that one a million times.
J. Paul Reed
At incident.io, agents now go beyond coding their mobile app changes, they actually run the app in a simulator and provide visual evidence of the change for a human to review.
Grayson Ho — incident.io
Getting Debezium to stream the first PostgreSQL change into Kafka is straightforward; keeping it correct through failover is the interesting part.
Frane Jelavic — ShiftMag
This one shows the drawbacks of focusing on a root cause through a really cool lens: a Five Whys analysis of the sinking of the Titanic.
Steve Mynett — Rootly
Giving an agent more context can make it less effective. Here's Meta's technique to mitigate that.
Plawan Rath, Akhil Jindal and Ashwini Verma — Meta
Migrating the busiest 0.00001% of repositories to Azure took some careful work and consideration of bottlenecks.
Eleanor White and Victoria Dye — GitHub
My favorite section is the one about achieving higher reliability by coordinating between their client library and the server side.
Matt Hammond — Ably
SRE Weekly, a production of Tinker Tinker Tinker, LLC · PO Box 253 · South Lancaster, MA 01561-0253 · USA